|
madsushi posted:Where are you getting that 100% number from? The guest OS? The guest OS usually can't tell what its actual CPU usage is and will often show 100% even when it's not consuming all of the CPU allocated to it. I'm going by the performance charts in vCenter. One thing that was weird is that the CPU is 2GHz, but it reported using around 2.3 GHz. Is that just a virtualization quirk?
|
#
?
Feb 23, 2014 05:49
|
|
|
|
| # ? Apr 25, 2024 16:49 |
|
|
Dr. Arbitrary posted:One thing that was weird is that the CPU is 2GHz, but it reported using around 2.3 GHz. Is that just a virtualization quirk? I think two causes; that is the actual processor usage to run the VM process so (vm allocated speed + overhead). Some Procs also have turbo boost features allowing for boosts above the advertised clocks.
|
|
#
?
Feb 23, 2014 06:02
|
|
|
Dr. Arbitrary posted:I'm going by the performance charts in vCenter. Intel turbo boost. It can clock individual cores higher than stock to boost single-threaded workloads. The fewer cores active, the more the individual core will boost. I haven't seen the turbo boost steppings listed on ARK but they're on wikipedia. For instance, an 8-core E5-2640 is 3/3/3/3/3/3/4/5. At 2 ghz base and 100 mhz Turbo boost increments, that processor will hit 2500 mhz if one core is loaded, 2400 if 2 cores, and 2300 if 3 or more cores are loaded. It's a feature to maximize performance while staying within the TDP.
|
|
#
?
Feb 23, 2014 06:22
|
|
|
TwoDeer posted:Happy to answer any questions you might have regarding Nutanix. Well I suppose the basic question is what's your experience? (and is this as an impartial end-user?) Ditto if anyone's used Simplivity as that is looking a better potential fit right now.
|
|
#
?
Feb 23, 2014 12:01
|
|
|
Bitch Stewie posted:Well I suppose the basic question is what's your experience? Edited my original response to indicate that I do work for Nutanix. If you want to PM me and let me know your region/vertical I'd be happy to put you in touch with reference-able users. I was in one of the very first partner training classes they had about two years ago and have been watching them ever since. Anyway, not interested in merely shilling for them but if there are specific questions I'll gladly take a stab at it. Curious as to why Simplivity appears to be a better fit, if you're able to elaborate.
|
|
#
?
Feb 23, 2014 16:50
|
|
|
TwoDeer posted:Edited my original response to indicate that I do work for Nutanix. If you want to PM me and let me know your region/vertical I'd be happy to put you in touch with reference-able users. I was in one of the very first partner training classes they had about two years ago and have been watching them ever since. Anyway, not interested in merely shilling for them but if there are specific questions I'll gladly take a stab at it. Curious as to why Simplivity appears to be a better fit, if you're able to elaborate. Disclosure appreciated  We've only just started speaking to a reseller so depending what comes back we'll be talking to Nutanix.I think what I'm trying to understand, and where I think Simplivity appears a better fit for us, is what you do if you essentially want a very basic 2 node cluster but want to stretch it out a bit? With Nutanix I think you're looking at a minimum of 4 nodes (2 "bricks" minimum of 2 nodes a brick so 8 processors?) which is a pretty big deal at SMB/SME size simply in vSphere/Windows licensing terms let alone hardware. Simplivity appear to need a minimum of 2 "cubes" and you'd think that if they're both on the same 1ms latency L2 subnet that shouldn't be an issue - again need to see what comes back to decide whether it's a waste of our time and theirs talking more. The reason for this is is a total environment refresh where I'm thinking if we're replacing servers (only 2 hosts right now) and SAN storage (approx 20TB usable capacity) and whatever we buy has to last 3 years, it's the right time to at least investigate converged. Regardless of vendor there is an appeal in simply buying a couple of "bricks" that do everything but equally there's caution about having to buy more computer capacity if all we need is a little more storage.
|
|
#
?
Feb 23, 2014 18:21
|
|
|
Bitch Stewie posted:The reason for this is is a total environment refresh where I'm thinking if we're replacing servers (only 2 hosts right now) and SAN storage (approx 20TB usable capacity) and whatever we buy has to last 3 years, it's the right time to at least investigate converged.
|
|
#
?
Feb 23, 2014 18:33
|
|
|
Bitch Stewie posted:Disclosure appreciated How does a 2-node cluster handle quorum? To that end, the smallest cluster currently supported from Nutanix consists of 3 nodes which would leave an expansion slot for most of our "blocks" (that's the term for the 2U appliance, a block - a block will have 1 or more "nodes [converged compute/storage]" in it). So the node becomes the unit at which you scale the size of your infrastructure. On physical footprint alone the minimum configuration for Nutanix is 2U with 3 nodes while Simplivity appears to require 4U for only two nodes. We support vSphere, KVM and Hyper-V for the hypervisor layer while I believe Simplivity currently only integrates with vSphere. However you decide to go, I think you're making the right move to look at converged infrastructure offerings.
|
|
#
?
Feb 23, 2014 20:38
|
|
|
TwoDeer posted:Edited my original response to indicate that I do work for Nutanix. Do you know Tim Federwitz? He runs http://www.virtuallygeeky.com/ He works for Nutanix, although he was at VMware a little over a year ago. Good guy. He really knows his VDI setups. He setup a 3D game demo at VMworld 2012 that was pretty cool.
|
|
#
?
Feb 24, 2014 01:39
|
|
|
DevNull posted:Do you know Tim Federwitz? He runs http://www.virtuallygeeky.com/ He works for Nutanix, although he was at VMware a little over a year ago. Good guy. He really knows his VDI setups. He setup a 3D game demo at VMworld 2012 that was pretty cool. Sure do; got to see the 3D gaming demo at VMworld 2012 and have been following his blog for a while. Tim gave me a hand in setting up a vDGA demo with one of our GPU nodes with 2 nvidia K2s for a customer. Agree that he is is a smart guy and a good dude.
|
|
#
?
Feb 24, 2014 05:37
|
|
|
madsushi posted:Where are you getting that 100% number from? The guest OS? The guest OS usually can't tell what its actual CPU usage is and will often show 100% even when it's not consuming all of the CPU allocated to it. Would that only be caused by resource scheduling or just a general rule of VMs?
|
|
#
?
Feb 24, 2014 11:27
|
|
|
TwoDeer posted:How does a 2-node cluster handle quorum? To that end, the smallest cluster currently supported from Nutanix consists of 3 nodes which would leave an expansion slot for most of our "blocks" (that's the term for the 2U appliance, a block - a block will have 1 or more "nodes [converged compute/storage]" in it). So the node becomes the unit at which you scale the size of your infrastructure. On physical footprint alone the minimum configuration for Nutanix is 2U with 3 nodes while Simplivity appears to require 4U for only two nodes. We support vSphere, KVM and Hyper-V for the hypervisor layer while I believe Simplivity currently only integrates with vSphere. Still waiting for a *lot* more info on both tbh. I know 2 node systems will handle split brain and if the system doesn't you'd hope there's a way to run a lightweight quorum as a VM on a Microserver or something so technically it shouldn't be an issue - but watch this space. I think the fact we'd need 6 nodes minimum Nutanix to do any kind of DR between rooms/locations has just ruled it out though simply on $$$ grounds based on real-world pricing I saw on a couple of 6020's.
|
|
#
?
Feb 24, 2014 13:53
|
|
|
I have a satellite office that has an old rear end IBM x3550 m2 running ESXi 4.0 with 4 vm's on its local storage. The dude that set up the server years ago used a 1mb block size for the datastore that all of the VM's are on which has a maximum file size of 256gb. The file server down there has over 1tb of data on a spanned volume of four 256gb disks... Veeam can't make a backup of the server because a snapshot is bigger than the 256gb limit. Is getting a small NAS or large external hard drive, downloading all of the datastore data to it, then reformatting the datastore with a larger block size then putting the datastore data back on a viable "fix" for this situation? What's out there that's relatively cheap that would give me some sort of data-protection? I don't particularly want to put the entirety of a server's production data on a single external hard drive. I had toyed with the idea of taking an "extra" host from our main office and loading it up with drives but drives aren't exactly cheap. Trying to find an alternative.
|
|
#
?
Feb 24, 2014 20:11
|
|
|
goobernoodles posted:I don't particularly want to put the entirety of a server's production data on a single external hard drive. Are you saying you don't have backups?
|
|
#
?
Feb 24, 2014 20:19
|
|
|
Wicaeed posted:Would that only be caused by resource scheduling or just a general rule of VMs? General rule of VMs, at least with VMware. I think someone said that if you have the latest Windows + latest VMware tools, it's better. But in general, the in-guest monitoring of stuff like CPU is really inaccurate. Just because you assign a guest 1 vCPU doesn't mean that it gets all of the cycles of that CPU all the time. VMware will only give it what it needs/asks for, so it can think it's at 100%. If VMware sees more instructions come in, it gives it more cycles, UP TO a total of 1 vCPU. The "up to" is the key here. VMware can just give you like 500 MHz of power when you're not doing anything and Windows can't tell the difference between that and being at 100% CPU.
|
|
#
?
Feb 24, 2014 20:34
|
|
|
I'm going to be taking a trip in a week or so time, are there any good books on Openstack that anyone could recommend as reading material during my flight?
|
|
#
?
Feb 24, 2014 22:04
|
|
|
Wicaeed posted:I'm going to be taking a trip in a week or so time, are there any good books on Openstack that anyone could recommend as reading material during my flight? You should bring a laptop, keep a few tabs open, and use packstack to install OpenStack inside a nested-virt VM. In short, no, there are no really good books on a product which has a new release every 6 months and is so new that those releases may completely revamp the recommended network configuration.
|
|
#
?
Feb 24, 2014 22:18
|
|
|
Wicaeed posted:I'm going to be taking a trip in a week or so time, are there any good books on Openstack that anyone could recommend as reading material during my flight? I'd be happy if there was more consise documentation on it. Sure it's gotten considerably better (not sure if that has anything to do with the vmware pledged support); but openstack is still far away from paper releases. goobernoodles posted:I have a satellite office that has an old rear end IBM x3550 m2 running ESXi 4.0 with 4 vm's on its local storage. The dude that set up the server years ago used a 1mb block size for the datastore that all of the VM's are on which has a maximum file size of 256gb. The file server down there has over 1tb of data on a spanned volume of four 256gb disks... Veeam can't make a backup of the server because a snapshot is bigger than the 256gb limit. hit me up on vent tonight I'll shoot the poo poo with you on some options
|
|
#
?
Feb 24, 2014 22:34
|
|
|
I'm trying to figure out a way to move the host management interface from one port group to another (in different vlans) on a VDS. Everything I've tried so far has rolled back. Has anyone pulled this off? Actual quote from VMware support: "just update the IP address in the DCUI and it'll automatically change port groups." Uh, what? KS fucked around with this message at 19:24 on Feb 25, 2014 |
|
#
?
Feb 25, 2014 19:01
|
|
|
KS posted:I'm trying to figure out a way to move the host management interface from one port group to another (in different vlans) on a VDS. Everything I've tried so far has rolled back. Has anyone pulled this off? Create a new one on the target then delete the old one. edit: optionally the migrate virtual machine networking wizard would work if the IP/subnet wasn't changing.
|
|
#
?
Feb 25, 2014 19:28
|
|
|
1000101 posted:Create a new one on the target then delete the old one. Surprisingly this doesn't work. I can create new vmnics all day but if I select the "use this interface for management traffic" option it loses connection to the host and rolls back. Is there only one interface at a time that gets that option or something? I haven't gone all-in and tried to change the default gateway at the same time. Guess I will.
|
|
#
?
Feb 25, 2014 23:48
|
|
|
What version of ESX?
|
|
#
?
Feb 25, 2014 23:50
|
|
|
5.1
|
|
#
?
Feb 26, 2014 00:00
|
|
|
Has anyone here successfully deployed vCenter Heartbeat in an existing vCenter 5.1u1 standalone environment? We have been having no end of hosed up problems since one of our the customer's "Technical Architects" deployed it. We opened a ticket with VMware about 2 weeks ago and have been troubleshooting with them but it really doesn't look like they can fix it. Honestly I'm just going to approach the TA in question and ask that he roll-back. On another note, what would you all consider an acceptable statistics logging level for a vCenter environment with 75 hosts and 600+ VMs? The aforementioned TA has set ours to Level 4 for the first and second intervals which in my opinion is amazingly retarded. We recently added 15 hosts to the PROD cluster and since then the stats-rollup SQL Agent jobs have been taking 180x longer to run. Honestly I think that Level 4 stats logging is crazy excessive however the TA in question says that they want "high-granularity" in VCOPS. Does stats Level 4 really expose any actually useful metrics?
|
|
#
?
Feb 27, 2014 18:45
|
|
|
How lovely would performance be if I tried to setup a iSCSI shared vmfs datastore for clustering on a disk array limited to two GB nics? It's a Synology RS814.
|
|
#
?
Feb 27, 2014 21:36
|
|
|
sudo rm -rf posted:How lovely would performance be if I tried to setup a iSCSI shared vmfs datastore for clustering on a disk array limited to two GB nics? It's a Synology RS814.
|
|
#
?
Feb 28, 2014 00:19
|
|
|
sudo rm -rf posted:How lovely would performance be if I tried to setup a iSCSI shared vmfs datastore for clustering on a disk array limited to two GB nics? It's a Synology RS814.  Guess where we went from 2 Lefthands with 2x1 GbE (Network RAID 10) to adding 2 4x1GbE nodes? This is 7 ESXi hosts (each with 2x1 GbE) and about 90 guests using shared storage. It's all about workload, it worked pretty well with light loads but lately some guests were crapping out due to high queue depth and latency (100s in loadavg with basically no CPU load)
|
|
#
?
Feb 28, 2014 00:23
|
|
|
That's some sweet graphing. How do you pull those stats?
|
|
#
?
Mar 1, 2014 03:55
|
|
|
What would you use if you typically used VSphere but wanted a zero cost solution for virtualization using 2 - 6 physical servers? migration capability and HA would still be required.
|
|
#
?
Mar 1, 2014 21:58
|
|
|
If you're just running Windows on it then a bunch of Server 2012 R2 Datacenter licenses and Hyper-V would be cheaper than VMware, but there's a cost involved in the form of System Center to manage it all properly. But it's a weird question because everything has a non-zero cost. If the issue is capex then run things in Azure rather than trying to hack something together.
|
|
#
?
Mar 1, 2014 22:06
|
|
|
dont change my name posted:What would you use if you typically used VSphere but wanted a zero cost solution for virtualization using 2 - 6 physical servers? migration capability and HA would still be required. If you are a startup, check out Bizspark. You get production licenses for Windows Server, System Center, and SQL server. Free of cost for several years or until you profit a set amount (think its like $1 million USD.) Oh and some free Azure capacity per month. Da Mott Man fucked around with this message at 22:16 on Mar 1, 2014 |
|
#
?
Mar 1, 2014 22:13
|
|
|
How do vmware licenses stack or whatever? I have vSphere 5 Enterprise and vCloud Suite Standard, along with a vCenter Server Standard license. (I'm a student, these are educational licenses). Right now I have my esxi box licensed as vcloud and vcenter. If I bring up another server, can I put the vsphere enterprise license on that and still manage it from vcenter? Or if I have to put a free esxi license on the 2nd machine can you still manage it from a licensed vcenter (I imagine you can.)
|
|
#
?
Mar 1, 2014 22:35
|
|
|
complex posted:That's some sweet graphing. How do you pull those stats? The data is from SNMP (LEFTHAND-NETWORKS-NSM-CLUSTERING-MIB::clusClusterStatsQDepthRead in this example). The graph is from Munin, I either wrote or found an SNMP plugin for Munin that grabbed cluster-wide read/write queue depth, bytes written/read and write/read latency. I'm not sure of the license on it, but the heavy lifting is: code:
|
|
#
?
Mar 2, 2014 00:03
|
|
|
dont change my name posted:What would you use if you typically used VSphere but wanted a zero cost solution for virtualization using 2 - 6 physical servers? migration capability and HA would still be required. Yeah the problem you are going to run into is that the 100% free alternatives all have major warts. That's how VMware can get away with charging a shitload, it all works and is comparatively simple to manage If you have Linux chops I'd suggest taking a look at Google's Ganeti project. It's certainly not as easy to operate as vSphere but it does the job if you have the expertise to set it up and write any supporting scripts/tools you need. For example, it can do live migration but (last I checked) it doesn't automatically do HA. You'd need your monitoring software to detect an instance had died and send a command to restart it on another node. As Caged said, you're not really going to get anything for free. Something like Ganeti has no licensing costs but you'll probably spend a lot in terms of man hours setting it up and maintaining it. But some managers don't care about that and only want to see $0 on their software expense report since hey, they were paying you already, it's "free" 
|
|
#
?
Mar 2, 2014 01:35
|
|
|
dont change my name posted:What would you use if you typically used VSphere but wanted a zero cost solution for virtualization using 2 - 6 physical servers? migration capability and HA would still be required. XenServer, it's free and works. Has some minor qwerks but it does work. http://www.xenserver.org/
|
|
#
?
Mar 2, 2014 01:45
|
|
|
I am pretty sure you can get most of the functionality of VMware + vsphere in XCP, or even the commercial xenserver offering now that it is free. There are some major missing features, which may not be relevant. 1) live storage migration (svMotion) 2) memory dedupe (page sharing) There are probably many others, but those are the big ones.
|
|
#
?
Mar 2, 2014 02:05
|
|
|
Is there a difference in the XenServer version on the XenServer.org website vs. the Citrix page?
|
|
#
?
Mar 2, 2014 02:38
|
|
|
adorai posted:I am pretty sure you can get most of the functionality of VMware + vsphere in XCP, or even the commercial xenserver offering now that it is free. There are some major missing features, which may not be relevant. oVirt has both of these features, plus HA and live migration.
|
|
#
?
Mar 2, 2014 06:14
|
|
|
What about the free Hyper V server 2012 r2? You won't be able to live migrate without AD though
|
|
#
?
Mar 2, 2014 06:50
|
|
|
|
| # ? Apr 25, 2024 16:49 |
|
|
Dilbert As gently caress posted:XenServer, it's free and works. Has some minor qwerks but it does work. Please don't use XenServer unless you really know your poo poo. I have been using XenServer for a pretty long time, probably about 6 years or so, and it has really gone downhill. It has more features, but it is buggy and unstable as gently caress. I have had nothing but issues with them since they released 6.1. I have spent a ton of hours this year troubleshooting serious XenServer stuff, all stemming from bugs in their product. In fact, I'm up at 6am on a Sunday restoring a VM from backup because another issue with XenServer. If you look at this thread or previous iterations, you'll see I recommended XenServer as an alternative if you can't afford both real VMware licenses and a SAN. At this point, VMware is too cheap and XenServer is too unstable. I would honestly look at Hyper-V before I considered another XenServer deployment.
|
|
#
?
Mar 2, 2014 13:01
|
|
